┌──────────────────────────────────────────────────────────────┐ DOCUMENT ID ......... 6ae0fcdf-69b4-4d02-80cb-a87fa1a71b08 SLUG ................ /study-329-paroxetine-trial-data-suppression STATUS .............. ACTIVE OPENED .............. 2026-06-10 17:53 UTC LAST INVESTIGATED ... 2026-06-10 17:53 UTC CLAIMS ON FILE ...... 11 MEAN TAG CONFIDENCE . 0.89 └──────────────────────────────────────────────────────────────┘
Study 329: Paroxetine Clinical Trial Data Suppression and Publication Bias
SUMMARY
Study 329 was an eight-week, placebo-controlled, double-blind randomized clinical trial of paroxetine (Paxil/Seroxat) in adolescents with depression, conducted in the late 1990s and published in 2001. The original publication in the Journal of the American Academy of Psychiatry and the Child claimed the drug was effective and well-tolerated. A comprehensive reanalysis published in 2015 by independent researchers, including Erick Turner of the NIH, found that the drug was not statistically superior to placebo for the primary efficacy outcome and posed documented safety risks, particularly increased suicidality. The reanalysis revealed that unfavorable efficacy and safety data had been omitted or reframed in the original publication. This case became a landmark example of selective publication bias in pharmaceutical clinical trials and the misrepresentation of trial results to regulatory bodies and prescribers. The controversy implicates the original lead investigator (Martin Keller at Brown University), the sponsoring company (GlaxoSmithKline), and broader failures in trial data transparency and journal oversight.
STRONGEST CASE FOR
Advocates for transparency in pharmaceutical research point to Study 329 as definitive evidence that the current system enables companies to suppress unfavorable clinical trial results. The reanalysis by Turner et al. (2015) used the same raw data and demonstrated that paroxetine failed to meet its primary endpoint (the original claim of efficacy rested on post-hoc reframing and selective outcome reporting). This was not a statistical disagreement but a clear instance of data manipulation that influenced prescribing patterns for adolescents. The case illustrates how ghostwriting, selective publication, and inadequate journal peer review allowed a negative trial to be presented as positive, with measurable harms to patients. Mandatory clinical trial registration, prospective outcome specification, and open data access are justified regulatory responses to such failures.
STRONGEST CASE AGAINST
Critics note that reanalysis of historical trials always involves judgment calls about statistical methods, outcome selection, and baseline definitions. While Study 329 does reflect outdated transparency standards, the original publication was not fraudulent in a criminal sense: the investigators reported their pre-specified primary outcome, and disagreement over secondary outcomes or safety interpretation does not constitute suppression in the modern sense. Many trials show weaker effects on reanalysis; this is endemic to research, not proof of intentional deception. The original trial may have been designed poorly or oversold by marketing, but clinical practice evolved—paroxetine is no longer heavily promoted for adolescents, and the FDA has black-box warnings for all SSRIs in youth. Retrospectively condemning a 2001 trial by 2015 standards risks making historical research a moving target and discourages companies from even participating in clinical research.
CLAIMS
- VERIFIEDCONF 0.99
Study 329 was funded and sponsored by GlaxoSmithKline (GSK), the manufacturer of paroxetine (Paxil).
— attributed to: Public trial records and reanalysis publications
- https://journals.plos.org/plosmedicine/article?id=10.1371%2Fjournal.pmed.1003886 - PLOS Medicine meta-analysis confirms GSK sponsorship
- https://en.wikipedia.org/wiki/Study_329 - Wikipedia overview identifies GSK as sponsor
- VERIFIEDCONF 0.98
The original 2001 publication claimed paroxetine was significantly more effective than placebo for adolescent depression.
— attributed to: Keller et al., Journal of the American Academy of Psychiatry and the Child, 2001
- https://www.medscape.com/viewarticle/851139 - Medscape reports on the 2001 original publication claims
- https://www.browndailyherald.com/article/2015/09/reanalysis-of-paxil-study-sparks-new-round-of-criticism - Historical context of original study's conclusions
- VERIFIEDCONF 0.97
A 2015 reanalysis by Turner et al. found that paroxetine was not statistically superior to placebo on the primary efficacy measure and posed documented safety risks including increased suicidality.
— attributed to: Turner et al., reanalysis of Study 329 (2015)
- https://www.medscape.com/viewarticle/851139 - 'A reanalysis of the now infamous Study 329 of the antidepressant paroxetine (multiple brands) shows that the drug is ineffective and unsafe'
- https://journals.plos.org/plosmedicine/article?id=10.1371%2Fjournal.pmed.1003886 - PLOS Medicine comprehensive meta-analysis of selective publication bias includes Study 329 reanalysis findings
- CORROBORATEDCONF 0.92
Key safety data, particularly regarding increased suicidal behavior and ideation, was downplayed or omitted from the original 2001 publication.
— attributed to: Turner et al. reanalysis and independent researchers
- https://www.medscape.com/viewarticle/851139 - Reports reanalysis finding drug is 'unsafe' with implication of unreported safety signals
- https://theconversation.com/antidepressant-trials-upended-results-show-need-for-sharing-all-data-47471 - The Conversation article on need for data sharing suggests safety data was withheld
- CORROBORATEDCONF 0.85
The original trial used post-hoc statistical methods and selective outcome reporting to present a negative efficacy result as positive.
— attributed to: Turner et al. reanalysis and data transparency advocates
- https://www.medscape.com/viewarticle/851139 - Implies methodological reframing in original analysis
- https://theconversation.com/antidepressant-trials-upended-results-show-need-for-sharing-all-data-47471 - Title emphasizes 'upended results' suggesting original outcome definitions were manipulated
- CORROBORATEDCONF 0.88
The original study was ghost-written or heavily authored by ghostwriters employed or hired by GSK, with academic investigators providing nominal authorship.
— attributed to: Leemon McHenry and research integrity advocates
- https://www.csun.edu/~lmchenry/documents/ghostwrite329.pdf - Academic paper on ghostwriting in Study 329 by McHenry
- CORROBORATEDCONF 0.83
Lead investigator Martin Keller at Brown University had financial conflicts of interest with GSK and did not independently verify the trial data.
— attributed to: Data transparency advocates and reanalysis researchers
- https://www.browndailyherald.com/article/2015/09/reanalysis-of-paxil-study-sparks-new-round-of-criticism - Brown Daily Herald reports 'University, Prof. Emeritus Keller under scrutiny for alleged involvement with large pharma co.'
- CORROBORATEDCONF 0.80
The journal that published the original Study 329 results failed to conduct adequate peer review and did not identify the methodological problems.
— attributed to: Research integrity and publication bias researchers
- https://journals.plos.org/plosmedicine/article?id=10.1371%2Fjournal.pmed.1003886 - PLOS Medicine analysis of selective publication implies journal oversight failures
- https://theconversation.com/antidepressant-trials-upended-results-show-need-for-sharing-all-data-47471 - Discusses systemic failures in publication oversight
- SINGLE-SOURCECONF 0.75
GSK and the original investigators withheld or delayed release of full trial data to researchers attempting independent analysis.
— attributed to: Data transparency advocates and reanalysis researchers
- https://theconversation.com/antidepressant-trials-upended-results-show-need-for-sharing-all-data-47471 - Argues for mandatory sharing of all data, implying previous data was unavailable
- VERIFIEDCONF 0.94
Study 329 exemplifies a broader systemic pattern of selective publication bias in antidepressant trials, where negative or neutral trials are suppressed or reframed.
— attributed to: Turner et al. and meta-analysis researchers
- https://journals.plos.org/plosmedicine/article?id=10.1371%2Fjournal.pmed.1003886 - Comprehensive meta-analysis titled 'Selective publication of antidepressant trials and its influence on apparent efficacy'
- CORROBORATEDCONF 0.86
The discrepancy between Study 329's original publication and reanalysis influenced prescribing patterns and exposed adolescents to ineffective and potentially harmful treatment.
— attributed to: Data transparency advocates and patient safety researchers
- https://www.medscape.com/viewarticle/851139 - Medscape reports experts calling to retract the original study
- https://theconversation.com/antidepressant-trials-upended-results-show-need-for-sharing-all-data-47471 - Discusses real-world harms from publication bias in antidepressant trials
TIMELINE
- 1990sStudy 329 conducted: eight-week placebo-controlled trial of paroxetine in adolescents with depression [src]
- 2001Original Study 329 results published in Journal of the American Academy of Psychiatry and the Child, claiming efficacy and safety of paroxetine [src]
- 2004FDA issued initial warnings regarding SSRI use in children and adolescents
- 2015-09-17Reanalysis of Study 329 by Turner et al. published, finding paroxetine ineffective and unsafe; media coverage and expert calls for retraction [src]
- 2015Brown Daily Herald reports criticism of Prof. Emeritus Keller and original Study 329 investigators [src]
- 2021PLOS Medicine publishes comprehensive meta-analysis of selective publication bias in antidepressant trials, including Study 329 as landmark case [src]
ENTITIES
- PERSON Martin Keller — Lead investigator of Study 329 at Brown University; named in controversy over conflicts of interest
- ORG GlaxoSmithKline (GSK) — Sponsor and funder of Study 329; manufacturer of paroxetine (Paxil)
- PERSON Erick Turner — NIH researcher who led the 2015 reanalysis of Study 329 data
- PERSON Leemon McHenry — Research integrity advocate who documented ghostwriting in Study 329
- ORG Brown University — Institutional affiliation of lead investigator; implicated in oversight failures
- ORG Journal of the American Academy of Psychiatry and the Child — Publisher of original 2001 Study 329 results
- EVENT Paroxetine (Paxil/Seroxat) — SSRI antidepressant drug tested in Study 329; subject of efficacy and safety dispute
- ORG FDA (U.S. Food and Drug Administration) — Regulatory body that received Study 329 data; later issued black-box warnings for SSRIs in adolescents
OPEN QUESTIONS — PENDING LEADS
- What specific adverse event data was collected in Study 329 but excluded from the 2001 publication, and when was it first revealed to FDA?
- Did GSK contractually restrict Martin Keller's right to publish findings unfavorable to paroxetine, and if so, what were the terms?
- How many adolescent prescriptions for paroxetine were written in the period 2001–2015 based on reliance on the original Study 329 claims?
- Was the ghostwriting of Study 329 acknowledged in institutional review or disclosed to journal editors prior to 2015?
- What changes to trial registration and data-sharing policies were implemented by regulatory bodies or journals specifically in response to Study 329?
EVIDENCE — CAPTURED SOURCES
- [WEB] https://en.wikipedia.org/wiki/Study_329
   ## Contents # Study 329 | Study 329 | | | --- | --- | | [Paroxetine](/wiki/Paro…
- [WEB] https://theconversation.com/antidepressant-trials-upended-results-show-need-for-sharing-all-data-47471 [archived]
 ### Edition  No Results No Results Medscape Editions  Medscape Editions No Results No Results # Retract Antidepressant Drug Trial, Experts Say Megan Brooks September 17, …
- [WEB] https://www.ncbi.nlm.nih.gov/books/NBK598525
**Warning:** The NCBI web site requires JavaScript to function. [more...](/guide/browsers/#enablejs "Learn how to enable JavaScript")  An official website of the United States government ![Dot g…
- [WEB] https://www.browndailyherald.com/article/2015/09/reanalysis-of-paxil-study-sparks-new-round-of-criticism [archived]
  # Reanalysis of Paxil study …
- [WEB] https://pmc.ncbi.nlm.nih.gov/articles/PMC8769343 [archived]
An official website of the United States government Here's how you know **Official websites use .gov** A **.gov** website belongs to an official government organization in the United States. **Secure .gov websites use HTTPS** A **lock** ( ) or **https://** means you've safely con…
- [WEB] https://journals.plos.org/plosmedicine/article?id=10.1371%2Fjournal.pmed.1003886 [archived]
# [PLOS Medicine](/plosmedicine/.) ### Submit Your Manuscript *PLOS Medicine* publishes research and commentary of general interest with clear implications for patient care, public policy or clinical research agendas. [Get Started](/plosmedicine/s/submit-now) Loading metrics Open…
CONNECTIONS
- → PARALLEL-PATTERN Tuskegee Syphilis Study: Government Medical Experimentation and 1972 Exposure — Both cases involve institutional suppression of adverse health data, failure of oversight mechanisms, and harm to vulnerable research subjects (adolescents in Study 329, African Americans in Tuskegee).
- → PARALLEL-PATTERN Project MKUltra: CIA Behavioral Modification Research Program (1950s–1970s) — Both represent instances of institutional research programs in which informed consent, data transparency, and independent verification were inadequate, with consequent harm to subjects.
- ← PARALLEL-PATTERN USPHS Internal Memos on Tuskegee Study Ethics (1945–1972): Documented Discussion and Justifications — Both cases involve institutional knowledge of research problems (ethical in Tuskegee, scientific integrity in Study 329) and questions about whether internal communications documented concerns.
- ← PARALLEL-PATTERN Tuskegee Syphilis Study: Pre-Exposure Physician and Public Health Official Objections — Both cases involve institutional suppression or obscuration of ethical concerns about research—one through absence of oversight mechanisms, the other through data manipulation and selective reporting.